Fix serialization on burst of unbinds - v2
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v7 Latest6 patches 1 replyIn review
- v66 patches 0 repliesSuperseded
- v56 patches 9 repliesSuperseded
Earlier versions · 4
| Version | Posted | Patches | Replies | Status |
|---|---|---|---|---|
| v4 | 5 patches | 4 replies | Superseded | |
| v3 | 2 patches | 1 reply | Superseded | |
| v2 | 2 patches | 0 replies | Superseded | |
| v1 | 1 patch | 10 replies | Superseded |
Patches in v7 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/6 | drm/xe: Enforce correct user fence signaling order usingdrivers/ drivers/ drivers/ +6 more | +86 | 0 replies | |
| 2/6 | drm/xe: Attach last fence to TLB invalidation job queuesdrivers/ drivers/ drivers/ +3 more | +156 | 0 replies | |
| 3/6 | drm/xe: Decouple bind queue last fence from TLB invalidationsdrivers/ drivers/ drivers/ +3 more | +143 | 0 replies | |
| 4/6 | drm/xe: Skip TLB invalidation waits in page fault bindsdrivers/ drivers/ | +13 | 0 replies | |
| 5/6 | drm/xe: Disallow input fences on zero batch execs and zero bindsdrivers/ | +36 | 0 replies | |
| 6/6 | drm/xe: Remove last fence dependency check from binds and execsdrivers/ drivers/ drivers/ +3 more | +0 | 0 replies |
Reviews trailers per patch
4 of 6 patches have a Reviewed-by or Acked-by.
| Patch | Matthew Brost | Thomas Hellström |
|---|---|---|
| 1/6 | Signed-of-by | R-b |
| 2/6 | R-b | |
| 3/6 | R-b | |
| 4/6 | ||
| 5/6 | ||
| 6/6 | R-b |
Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.
Cover letter [PATCH v7 0/6] Fix serialization on burst of unbinds - v2
Attempting to resolve part of [1]; this solution differs than v1 [2] by changing last fence semantics detailed in the patches.
Overview of issue in [1]:
When a burst of unbind jobs is issued, a dependency chain can form between the TLB invalidation of a previous unbind job and the current one. This leads to undesirable serialization, causing current jobs to wait unnecessarily for prior TLB invalidations, execute on the GPU when not needed, and significantly slow down the unbind burst—resulting in up to a 4× to 8× slowdown depending on platform.
Matt
[1] https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/6047
[2] https://patchwork.freedesktop.org/series/156144/
Matthew Brost (6):
drm/xe: Enforce correct user fence signaling order using drm/xe: Attach last fence to TLB invalidation job queues drm/xe: Decouple bind queue last fence from TLB invalidations drm/xe: Skip TLB invalidation waits in page fault binds drm/xe: Disallow input fences on zero batch execs and zero binds drm/xe: Remove last fence dependency check from binds and execs
drivers/gpu/drm/xe/xe_exec.c | 7 +- drivers/gpu/drm/xe/xe_exec_queue.c | 120 ++++++++++++++++++++--- drivers/gpu/drm/xe/xe_exec_queue.h | 23 ++++- drivers/gpu/drm/xe/xe_exec_queue_types.h | 12 +++ drivers/gpu/drm/xe/xe_migrate.c | 14 +++ drivers/gpu/drm/xe/xe_migrate.h | 8 ++ drivers/gpu/drm/xe/xe_oa.c | 45 ++++++--- drivers/gpu/drm/xe/xe_oa_types.h | 8 ++ drivers/gpu/drm/xe/xe_pt.c | 80 +++++---------- drivers/gpu/drm/xe/xe_sched_job.c | 17 ---- drivers/gpu/drm/xe/xe_sched_job.h | 1 - drivers/gpu/drm/xe/xe_sync.c | 91 ++++++++++------- drivers/gpu/drm/xe/xe_sync.h | 3 + drivers/gpu/drm/xe/xe_sync_types.h | 3 + drivers/gpu/drm/xe/xe_tlb_inval_job.c | 31 ++++-- drivers/gpu/drm/xe/xe_tlb_inval_job.h | 5 +- drivers/gpu/drm/xe/xe_vm.c | 97 +++++++++++------- drivers/gpu/drm/xe/xe_vm_types.h | 6 +- 18 files changed, 377 insertions(+), 194 deletions(-)
Signature · 1 line
2.34.1